1. Fabienne Fortanier Head of Trade Statistics OECD
Accounting for heterogeneity:
Possibilities of using TEC to extend supply & use tables
Fabienne Fortanier
Head of Trade Statistics
OECD

2. Accounting for heterogeneity
Key assumption in TiVA: homogeneous industries
all firms allocated to a particular industry have the same ‘ production’ function
That for a given product, imports by industries (firms) are sourced from the same mix of countries
Key challenge: taking the GVC beyond the industry level: accounting for firm heterogeneity
Not all firms in each industry are equally (or similarly) involved in GVCs
Extent of GVC involvement is correlated with virtually all dimensions of GVC impact, including value added, productivity, technology, size, growth, survival rate

3. Accounting for heterogeneity
To what extent can we capitalize on existing statistics, including TEC, STEC, FATS, “TEC+”, to
a) improve our estimates (of e.g. the import content of exports) and
b) develop new, ‘richer’ indicators on GVCs (how, where, who)
These statistics are policy relevant in their own right and for integration within TiVA framework

4. This presentation
Highlights the importance of heterogeneity (using TEC)
Explains more about TEC & how the data are compiled
Lists the ‘ideal’ additional data needs for breaking down SUTs by enterprise characteristics
Requires combination of TEC and additional business statistics (including SBS and FATS)
Illustrates its possible uses in breaking down supply and use tables and challenges to overcome
Gives an overview of current steps taken by the OECD to further this research agenda

5. Trade is concentrated among a few enterprises
(% of total* export accounted for by top # enterprises)
* For EU countries, data refer to extra-EU exports instead of total

6. Exporting firms are responsible for the majority of imports (1)
Unfortunately I have this import value-export status link only at the country level in TEC ‘Trade by Type of Trader ’, so it is not easy to exclude wholesalers

7. Exporting firms are responsible for the majority of imports (2)
The value of imports per firm are on average 20 times greater for exporters than for non-exporters
Values in 1000 euros

8. Trade is concentrated among large firms
Size matters (2)
Trade is concentrated among large firms
(export value by firm size class in employees)

9. Very few firms are foreign owned, but…
Ownership matters (1)
Very few firms are foreign owned, but…
(foreign owned firms and foreign affiliates, as % of all enterprises)

10. Ownership matters (2)
foreign owned enterprises account for ~ 1/3 of turnover, 1/5 of employment and 1/4 of value added in OECD countries…
(% of turnover, employment and value added by foreign owned enterprises)

11. …and for at least 1/3 of international trade
Ownership matters (3)
…and for at least 1/3 of international trade

12. Trade by Enterprise Characteristics (TEC)
All these presented examples are based on TEC data (OECD/Eurostat):
Trade by Size Classes & Economic Activity
Trade by Economic Activity & Top Enterprises
Trade by Economic Activity & Partner Zones
Trade by Economic Activity & Number of Partner Countries
Trade by Economic Activity & Commodity Group
Trade by Type of Ownership & Economic Activity
Domestically controlled enterprises (with or without own affiliates)
Foreign controlled enterprises

13. TEC coverage and compilation
About the OECD-Eurostat TEC database:
Covers 33 countries (6 non-OECD) from 2005 onwards, up to 2 digit level of ISIC rev 3 & 4
Based on trade statistics linked to the business register
Matching rates excellent to perfect
But some issues remain
e.g. non-resident trade
overemphasis of wholesalers
Confidential data

14. Use Table

15. Accounting for heterogeneity in SUTs
A strongly simplified example (1 industry)
Homogeneous
(now)
Heterogeneous
(future)
All Industry
‘Domestic’
‘Trader’
Imports 25
Domestic purchases 30 15
Value added 45
Total output
100 70
Export 60
Domestic sales 40 10
Import content of export:
25%
(25 / 100)
36%
(25 / 70)

16. Information needs
For a breakdown between exporting / non exporting firms:
Geographical breakdown (e.g. main trading partners, regions) of imports and exports (by exporting / non-exporting firms)
Purchases, output, value added (employment if possible), by industry, broken down by exporting vs non-exporting firm
 link TEC with SBS
For a breakdown with (also) foreign ownership:
(Same as above) + exports by ownership, industry and foreign ownership
 link TEC with FATS

17. Feasibility of ‘macro-linking’
Use the information from the three different data sources at the ISIC 2-digit industry level
Illustrates (again) the importance and relevance of accounting for heterogeneity
BUT: highlights also some problems that can only be solved at micro level

18. Examples of ‘Macro-linking’ (1)
Foreign-controlled enterprises are more export intensive
Export to turnover ratios, total economy (2011)
the estimates are obtained by linking turnover information from FATS to import and export information from TEC, using common industry identifier at two-digit level as linking information. For Italy, data on export and imports are also obtained from FATS statistics. The estimates are available at the two-digit industry level.

19. Examples of ‘macro-linking’ (2)
Value added over turnover ratio tends to be higher for domestic firms
Value added/Turnover by firm ownership, total economy (2011)

20. Challenges (1): wholesale is treated very differently across data sources
Imported intermediates / total intermediates: comparing SU table, proportional allocation and TEC data (Italy 2010 )
Here we show the proportion of imported inputs over total inputs as measured from the national S/U table (blue), splitting the total I/0 table using the proportionality assumption (industry’s imports of each commodity, relative to its total demand, is the same as the economy-wide imports relative to total demand for the commodity, in red), and using TEC data (in green). Intermediates in TEC are estimated using the fraction of intermediate over total use by commodity groups from national S/U table. You can see that TEC performs less well than the proportional allocation by comparing the correlations with the ‘true’ shares based on import flow matrix. This is mostly due to the issue of wholesale imports in TEC.
Correlation S/U and proportional allocation: 0.8
Correlation S/U and TEC intermediate imports: 0.55

21. Challenges (2): Populations are inconsistent across data sources
Number of 2 digit industries where the number of exporters in TEC is larger than the number of enterprises in SBS, by size, 2011

22. Challenges (3): Variable values are inconsistent across data sources
Number of industries for which exports (from TEC) are larger than turnover (SBS) (by firm size, 2011)

23. Challenge (4): Inconsistencies between TEC and FATS
Differences in the share of imports of foreign-controlled firms as measured in TEC and in FATS (Italy 2011)
Here there seems to be a clear story on unidentified domestic trade in TEC…
Pharmaceutical products

24. Current activities at the OECD
Increase country coverage in TEC data
Investigate methodological improvements
Better explanation of matching rates (non-resident trade);
Better assignment of trade accounted for by retailers and wholesalers to underlying economic activities;
Explore if confidentiality issues can be overcome by deciding on (harmonized) aggregation of certain items;
Include services (STEC)
Add to visibility: e.g. website on TEC : with country notes (forthcoming)
Continue with feasibility studies re. heterogeneity in SUTs (including work with individual countries who can provide data)
Institutional: WPTGS; Expert Group on Extended SUTs

25. Thank you Contact: Fabienne Fortanier fabienne.fortanier@oecd.org
Head of Trade Statistics
TEC :
TiVA :